Scholarlink

For the issue of low detection accuracy due to the large variation in the shape and size of defects on the surface of hot strip steel, an improved detection algorithm based on YOLOv8 is proposed. LSKNet is used to replace the original backbone network, and Dual Attention Block attention mechanism and Strip Block strip convolution module are introduced to enhance the feature extraction capability. Experiments show that the algorithm achieves a mAP of 77.5% on the NEU-DET dataset, which is 4.8% higher than the original algorithm.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improved Algorithm for Hot-Rolled Steel Strip Surface Defect Detection Based on YOLOv8

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hongbo Li,
  • Xiaobo Jiang,
  • Xiaobin Zhang,
  • Yuxin Li,
  • Yuhui Li

摘要

For the issue of low detection accuracy due to the large variation in the shape and size of defects on the surface of hot strip steel, an improved detection algorithm based on YOLOv8 is proposed. LSKNet is used to replace the original backbone network, and Dual Attention Block attention mechanism and Strip Block strip convolution module are introduced to enhance the feature extraction capability. Experiments show that the algorithm achieves a mAP of 77.5% on the NEU-DET dataset, which is 4.8% higher than the original algorithm.